www.elsevier.nl/locate/carres

Carbohydrate Research 330 (2001) 93–101

Polysaccharides of green Arabica and Robusta coffeebeans

Monica Fischer,* Silvia Reimann, Veronique Trovato, Robert J. RedgwellNestle Research Center, Vers-Chez-les-Blancs, CH-1000 Lausanne 26, Switzerland

Received 11 July 2000; received in revised form 12 October 2000; accepted 13 October 2000

Abstract

Two independent procedures for the quantitative determination of the polysaccharide content of Arabica Caturra(Coffea arabica var. Caturra) and Robusta ROM (Coffea canephora var. ROM) green coffee beans showed that theyboth contained identical amounts of polysaccharide. Cell wall material (CWM) was prepared from the beans andpartial solubilisation of component polysaccharides was effected by sequential extraction with water, 1 M KOH,0.3% NaClO2, 4 M KOH and 8 M KOH. The monosaccharide compositions of the CWMs were similar, althoughArabica beans contained slightly more mannose than Robusta. In the latter, more arabinogalactan was solubilisedduring preparation of the CWM and the water-soluble fraction of the CWM contained higher amounts ofgalactomannan than in Arabica. Linkage analysis indicated that the galactomannans possessed unbranched tobranched mannose ratios between 14:1 and 30:1 which is higher than previously reported. No major difference in thestructural features of the galactomannans between species was found. The arabinogalactans were heterogeneous bothwith regard to the degree of branching and the degree of polymerisation of their arabinan side-chains. Compared toArabica, Robusta appeared to contain greater amounts of arabinogalactans with longer side chains. It is concludedthat there was no detectable difference between the Arabica and Robusta varieties of this study in their absolutepolysaccharide content or in the gross structural features of their galactomannans. Differences were apparent bothin the structural features and ease of solubility of the arabinogalactans but a more detailed study of several varietiesof Arabica and Robusta will be required to determine whether these differences occur consistently between species.© 2001 Elsevier Science Ltd. All rights reserved.

Keywords: Cell walls; Polysaccharides; Coffee beans

1. Introduction

Cell wall polysaccharides constitute half thedry weight of the native coffee bean but thereis still much to be revealed about the detailedstructure of individual polysaccharides andhow they interact to give the integrated struc-ture of the cell wall. Thaler1,2 and Wolfrom3–6

detected the presence of mannans, galac-tomannans, arabinogalactans and cellulose asthe main polysaccharides in green and roastedcoffee beans. A more recent study7,8 has re-ported detailed structural data on the twoprincipal non-cellulosic polysaccharides. Themannan purified by Bradbury was composedof (1�4)-linked b-mannan chains substitutedat O-6 with single galactose residues approxi-mately every 100 residues.

The arabinogalactan isolated by these au-thors had an arabinose/galactose ratio of 0.4/1

* Corresponding author. Tel.: +41-21-7858702; fax: +41-21-7858554.

E-mail address: monica.fischer@rdls.nestle.com (M. Fis-cher).

0008-6215/01/$ - see front matter © 2001 Elsevier Science Ltd. All rights reserved.

PII: S 0008 -6215 (00 )00272 -X

M. Fischer et al. / Carbohydrate Research 330 (2001) 93–10194

and consisted of a backbone of (1�3)-linkedb-galactose substituted at O-6 with arabinoseand/or galactose residues. The side-chainscontain arabinose and galactose residues witharabinose as terminal residue. These linkagesare characteristic of type-II arabinogalactans,a polymer which is usually covalently linkedto protein.

Some more recent work9 has focussed onthe polysaccharides from hot water extracts ofroasted Arabica beans. An arabinogalactanand a mannan with similar structural featuresto the polymers described above were isolateddespite the fact that structural modificationswere induced by roasting.

Differences in total polysaccharide contentas well as structural differences between Ara-bica and Robusta species have been discussedin the literature. Reliable values are difficultto obtain as they generally have to be extra-polated and recalculated from a series of val-ues obtained after various treatments and ex-tractions which are not comparable.8,10–12

Most of these sources report a lower polysac-charide content for Robusta compared toArabica. Clifford11 reported a total polysac-charide content between 38 and 48% for Ro-busta and 48–55% for Arabica, butexperimental information detailing how thevalues were obtained is lacking. In contrast,Bradbury and Halliday8 report a totalpolysaccharide content of 48.1% for green Ro-busta coffee from the Ivory Coast and ‘alower polymeric carbohydrate content forArabica beans’.

Clifford11 also reports that Arabicas maycontain more arabinogalactan (9–13%) thanRobusta (6–8%) and more galactomannan(25–30% vs. 19–22%). He suggests that thegalactomannan in Robusta is more highlybranched and thus less crystalline but thesespeculations are not well substantiated. This isalso used to explain why, for a same degree ofroasting, Robustas generally produce moresoluble solids than Arabicas.10–12

The present study compares the totalpolysaccharide content of a variety of Arabicaand of Robusta and report data for the struc-tural features of arabinogalactan and galac-tomannan isolated from each.

2. Materials and methods

Raw material.—Dry, depulped and de-parched beans of Arabica Caturra (Coffeaarabica var. Caturra) and Robusta ROM(Coffea canephora var. ROM) coffee whichwere harvested at full maturity in Ecuadorwere used for this investigation.

General.—Polysaccharide fractions and cellwall material (CWM) were hydrolysed with 2M TFA for 1 h at 110 °C and converted toalditol acetates for GLC analysis as describedin Ref. 14. Insoluble materials were hy-drolysed using Saeman hydrolysis with 72%H2SO4 at room temperature for 3 h thendilution to 1 M H2SO4 and hydrolysed at105 °C for 2 h. The samples were thenanalysed by HPAE-PED liquid chromato-graphy.

Polysaccharides were methylated using amodification of the method of Ciucanu andKerek.13 GLC–MS of the partially methylatedalditol acetates was accomplished as describedpreviously.14

Isolation and fractionation of cell wall mate-rial.—Triplicate samples were taken of eachspecies of beans. Unless specified, all opera-tions were performed at ambient temperature.

CWM was prepared and fractionated usingprocedures adapted from existing methodsused for fruit tissue.15 Cryo-milled (IKA-Uni-versalmuhle, Staufen, Germany) coffee beanswere homogenised in two volumes of cold(4 °C) (2:1:1 w/v/v) phenol–AcOH–water(PAW). The PAW-soluble fractions were com-bined, dialysed and the polymers recoveredafter freeze-drying. The residue, hereaftercalled CWM, was suspended in water anddialysed for 2 days at 4 °C then freeze-dried.

Before attempting to fractionate thepolysaccharides, the CWM was defatted bystirring 1 g overnight in 100 mL ofdichloromethane and air-drying. DefattedCWM (1 g) was sequentially extracted with100 mL of water (2 h), 1 M KOH (2 h),NaClO2 (0.3 g/100 mL+0.12 mL AcOH, un-der argon, 2 h, 70 °C), 4 M KOH (2 h), 8 MKOH (2 h). All alkaline solvents contained 20mM NaBH4. Solubilised polymers were recov-ered following neutralisation of the solution topH 5.0, dialysis and freeze-drying.

M. Fischer et al. / Carbohydrate Research 330 (2001) 93–101 95

Table 1Yield of dichloromethane-soluble, PAW-soluble, PAW-pre-cipitate and CWM fractions isolated during CWM prepara-tion from Arabica Caturra and Robusta ROM coffee beans(mean values; n=3)

Fraction Amount (g per 100 g fwt)

ArabicaRobusta

Dichloromethane 9.4 9.7PAW-soluble 1.9 4.8

trace3.1PAW-precipitate51.7CWM 53.7

min. The samples were centrifuged and theresidue was stirred for 2 h in 100 mL of 80%EtOH at ambient temperature. The superna-tants were combined and dried down. Theresidues were suspended in water and dialysedfor 2 days in a membrane with a molecularweight cut-off of 3.5 kDa. The entire contents(residue and supernatant) of the dialysis bagwere freeze-dried to give the total polysaccha-ride. The carbohydrate content of the totalpolysaccharide fraction was then determinedby two procedures. By the phenol–sulfuricassay and by GLC analysis of the alditolacetates after Saeman hydrolysis of thepolysaccharide.

3. Results

Preparation of CWM.—CWM was pre-pared by a modification of an existing methodused for the isolation of CWM from fruittissue.15 PAW was used to inactivate endoge-nous enzyme activity and solubilise intra-cel-lular proteins which have a tendency to bindto the cell wall during its isolation. The CWM(Table 1) from Arabica and Robusta beansaccounted for 53 and 52% of the startingmaterial, respectively.

Composition of CWM.—The monosaccha-ride composition of the CWM was determinedfollowing Saeman hydrolysis (Table 2) andwas essentially the same for Arabica and Ro-busta. However, there was the suggestion thatArabica contained slightly more mannosethan Robusta. Mannose was the most abun-dant monosaccharide followed by galactose,glucose and arabinose.

Composition of other fractions isolated fromthe coffee beans.—The PAW extract separatedinto a soluble and insoluble fraction during

Size-exclusion chromatography.—A columnof Sephacryl S-300 (100 cm×16 mm) wasused at a flow rate of 25 mL/h with a sodiumacetate buffer, 0.05 M, pH 6 containing 125mM NaCl. The column was calibrated usingdextran standards.

Samples corresponding to approximately 20mg of carbohydrate material were dissolved in1 mL of acetate buffer. Fractions were col-lected and analysed for total sugars using thephenol–H2SO4 test. Selected fractions werepooled, freeze-dried and used for monosac-charide determination and methylationanalysis.

TCA precipitation of the PAW-soluble mate-rial.—Water (2 mL) was added to the PAW-soluble fraction (100 mg) followed by 2 mL oftrichloroacetic acid 0.4 M (TCA). The sampleswere left for 2 h at room temperature, thencentrifuged for 10 min at 3000 rpm. The su-pernatant was dialysed with a membrane cut-off of 3.5 kDa. The freeze-dried supernatantwas then partially methylated and acetylatedfor linkage analysis.

Total polysaccharide content.—Green beanswere cryo-milled to a fine powder and 10 gwere refluxed in 100 mL of 80% EtOH for 15

Table 2Monosaccharide composition of CWM following Saeman hydrolysis (mean values; n=3) a

Total (mg/mg)Monosaccharide composition (mol%)Sample

GlcGalManXylAraFucRha

25.5 17.6 6780.4 10.8 0.6Robusta 44.80.324.6 16.8 695traceArabica 9.90.3 0.8 47.5

a Protein content: 13.4 and 12.5% for Arabica and Robusta beans, respectively.

M. Fischer et al. / Carbohydrate Research 330 (2001) 93–10196

Table 3Monosaccharide composition of PAW-soluble and PAW-precipitate fractions isolated during the purification of CWM fromArabica and Robusta coffee

Monosaccharide (mol%) Total (mg/mg)

Fuc Ara Xyl Man GalRha Glc

Robusta0.4 37.9 2.3 7.0 38.7 1.3 402PAW-soluble 12.42.4 21.7 3.1 19.2 26.18.7 18.8PAW-prec. 27

Arabica0.2 33.2 3.0 11.2 40.5 5.8PAW-soluble 1216.0

Table 4Monosaccharide composition and total polysaccharide content of green Arabica and Robusta coffee (mean values9SE, n=3)

Rha (mol%) Ara (mol%)Sample Gal (mol%)Fuc (mol%) Glc (mol%) Xyl (mol%) Man (mol%) Total(ug/mg)

traceRobusta 1.090.1 12.290.4 26.490.1 16.390.5 1.090.2 43.191.1 55591.90.490.0 10.690.1 24.590.1 16.790.1trace 1.190.1Arabica 46.790.2 55891.3

dialysis. In Robusta, the PAW-soluble frac-tion (Table 3) consisted of 40% polysaccharidemost of which was arabinogalactan. The ma-terial that precipitated during dialysis con-tained very little polysaccharide and waspredominantly protein (data not shown).PAW solubilised more polysaccharides fromRobusta than from Arabica.

Total polysaccharide content.—The totalpolysaccharide content of the beans was mea-sured after removing the low molecular weightcarbohydrates by ethanolic extraction. Theethanol-insoluble material (total polysaccha-ride) was dialysed to remove the last traces ofmonosaccharide and recovered by freeze-dry-ing. Total carbohydrate as determined by thephenol–sulfuric assay gave identical yields ofpolysaccharide in Arabica and Robusta. Thiswas supported by GLC analysis of the alditolacetates following Saeman hydrolysis of thetotal polysaccharide (Table 4). Thus, in con-trast to previous reports,7,8,10–12 we found noevidence for differences between Arabica andRobusta in their total polysaccharide content.

Composition of cell wall polysaccharide frac-tions.—The heterogeneous nature of the cellwall means that differences in the relativeamounts and sugar compositions of individualpolysaccharides may exist without any appar-

ent difference in the monosaccharide composi-tion of the CWM as a whole. Therefore,individual polysaccharide fractions were iso-lated by sequential extraction of CWM withwater, 1 M KOH, NaClO2, 4 M KOH and 8M KOH and their monosaccharide composi-tion determined (Table 5).

The major variation in the yield of an indi-vidual fraction occurred in the water-solublefraction which had a higher yield in Robusta.The mannose content of this fraction washigh. Close to 10% of the total mannose fromRobusta beans was solubilised in the waterfraction compared to less than 1% fromArabica.

Although there were significant differencesin the comparative amounts of the NaClO2-, 4M KOH- and 8 M KOH-soluble fractions,these fractions accounted for only a small partof the total polysaccharide.

Acidified sodium chlorite is traditionallyused for delignification but has also beenshown to be a good solvent for glycoproteins16

and it proved to be a good solvent for ara-binogalactans. Increasing concentrations ofKOH solubilised a mixture of poly-saccharides.

The composition of the residue which ac-counted for 60 and 77% of the CWM of

M. Fischer et al. / Carbohydrate Research 330 (2001) 93–101 97

Table 5Yield and monosaccharide composition of fractions of CWM isolated from Arabica and Robusta beans (mean values; n=3)

Monosaccharide composition (mol%)% CWM Total (mg/mg)Fraction

Rha Fuc Ara Xyl Man Gal Glc

Water-soluble1.5 0.4 14.7Robusta 1.415.6 36.8 29.3 15.9 537

3.5Arabica 3.2 0.2 25.9 1.3 16.7 49.3 3.4 273

1 M KOH-soluble1.8 0.4 17.3 2.3 35.1Robusta 33.89.1 9.2 4042.4 0.8 20.5 2.7 35.112.8 35.3Arabica 3.2 149

NaClO2-soluble1.6 0.7 32.4 0.54.5 5.1Robusta 57.8 1.9 3621.3 0.3 33.3 1.0 2.4 60.7Arabica 1.02.4 337

4 M KOH-soluble0.5 0.4 14.3Robusta 4.44.4 33.6 34.1 12.6 485

2.1Arabica 0.6 0.2 17.7 13.9 29.7 31.4 6.5 345

8 M KOH-solubleRobusta 5.5 0.4 0.1 23.8 1.6 17.1 53.2 3.8 312

1.1 0.2 25.9 4.3 11.6 54.62.6 2.4Arabica 361

Residue0.1 nd. 8.6 0.1 46.8 26.3 18.0 698Robusta 61.10.7 0.4 7.0 3.376.6 51.3Arabica 19.0 18.3 701

Robusta and Arabica respectively, resembledthat of the original CWM, demonstrating thedifficulty of solubilising not only the galac-tomannan but also the arabinogalactan in cof-fee CWM.

Linkage analysis of PAW-soluble arabino-galactan.—The arabinogalactans recovered inthe PAW-soluble fractions were deproteinatedby TCA precipitation, methylated, convertedinto partially methylated alditol acetates andexamined by GLC–MS (Table 6).

The arabinose/galactose ratio was close tounity for both species as was the ratio of3-linked to 3,6-linked galactose (0.9:1 for Ro-busta, 1.2:1 for Arabica). Arabinogalactancontained amounts of 5-linked arabinose indi-cating that the side-chains consisted of morethan a single arabinose unit. The presence ofterminal galactose indicated that some of theside-chains contained galactose as chain end-ing residues.7

Characterisation of indi6idual polysaccha-rides.—Selected CWM fractions were purifiedfurther by subjecting them to size-exclusionchromatography (SEC) on Sephacryl S-300 inpreparation for structural characterisation by

linkage analysis. During SEC, the arabino-galactans generally eluted at a higher MWthan the galactomannans (Fig. 1, Table 7),enabling recovery of the polysaccharides in amore purified form for methylation analysis.Pooled polysaccharide fractions recovered af-ter SEC were methylated, converted into par-tially methylated alditol acetates andexamined by GLC–MS.

Structural features of coffee arabinogalac-tan.—The glycosyl linkages in Table 8 areconsistent with the documented structure of

Table 6Linkage analysis of polysaccharide content of PAW-solublefraction

RobustaLinkage Arabica

3.77.4terminal-Rha22.1terminal-Araf 22.8

5-Araf 13.711.66.5 aterminal-Xylp

terminal-Gal 2.7 3.115.33-Gal 19.5

3,6-Gal 16.815.85.54-Man 7.1

a Not detected.

M. Fischer et al. / Carbohydrate Research 330 (2001) 93–10198

Fig. 1. Gel-permeation chromatography on Sephacryl HR-300 of water-soluble and NaClO2-soluble fractions from Arabica(triangles) and Robusta (squares) beans.

an arabinogalactan which possesses a (1�3)-linked galactose backbone with a proportionof the galactose residues substituted at O-6 byside-chains of arabinose and/or galactoseresidues. The degree of substitution as indi-cated by the ratio of 3:3,6-linked galactoseranged between 0.92 (Robusta Chlorite-solu-ble) and 1.5 (Robusta 8 M KOH-soluble). Asimilar range of degree of branching wasshown by the arabinogalactan in the Arabicafractions, however the pattern of distributionof the branched arabinogalactans differed be-tween Robusta and Arabica. For example, thewater-soluble arabinogalactan in Robusta wasmore highly branched than the equivalentfraction in Arabica and these differences werealso apparent in the arabinogalactans fromthe sodium chlorite and 8 M KOH-solublefractions. The PAW-soluble fraction of Ro-busta contained greater amounts of arabino-galactan than the same fraction from Arabica.However, linkage analysis of these arabino-galactans did not reveal any marked differ-ences in their structural features (Table 6).

It was noticeable that the ratio of 5-linkedarabinose to terminal arabinose also showedconsiderable variation among fractions. Ahigher proportion of 5-linked residues wouldindicate that the side chains of the arabino-galactans were more extended than thosepolymers with higher proportions of terminalarabinose. In general it seemed that the moreextended the side-chains were the more easilythe arabinogalactans were removed from theCWM. Thus, in Robusta, the ratio of terminalto 5-linked arabinose was highest in the water-soluble and 1 M KOH soluble fractions and

lowest in the 4 M and 8 M KOH-solublefractions. Arabica exhibited a similar pattern,although the proportion of 5-linked arabinosewas in general lower than in the same Robustafractions.

Structural features of coffee galactomannans(Table 9).—Galactomannans were recoveredin most of the CWM fractions, although inRobusta coffee a higher proportion was recov-ered in the water-soluble fraction. After thefinal extraction with 8 M KOH, the residuestill contained close to 50% of mannose. Thewater-soluble galactomannans had ratios ofunbranched to branched mannose of 19:1 forRobusta and 29:1 for Arabica. In the othercell wall fractions, the material was morebranched with ratios of unbranched tobranched residues close to 10:1, except in the 4M KOH fractions where the galactomannanswere less substituted (20:1).

Table 7Distribution of the main monosaccharides (mol%) in thefractions recovered after size-exclusion chromatography ofthe water and NaClO2-soluble fractions of green Arabica andRobusta CWM

Ara Gal Glc Man

Water-soluble fraction5.449.7 8.629.3F1Robusta

2.31.657.032.2F1ArabicaF2 8.2Robusta 14.3 19.1 47.3

Arabica 46.3F2 19.0 14.9 5.3

NaClO2-soluble fraction1.7 2.0Robusta F1 31.3 58.8

0.52.654.9Arabica 38.2F1F2 36.9 9.2 5.7Robusta 33.2

13.73.118.747.2Arabica F2

M. Fischer et al. / Carbohydrate Research 330 (2001) 93–101 99

Table 8Glycosyl-linkage analysis of the arabinogalactan-rich polysaccharide fractions from Arabica and Robusta coffee

1 M KOH-soluble NaClO2-soluble 4 M KOH-solubleFraction 8 M KOH-solubleWater-soluble

Arabica Robusta Arabica Robusta ArabicaLinkage RobustaRobusta Arabica Robusta Arabica

terminal-Araf 18.116.0 16.0 14.8 22.8 15.1 21.5 16.3 17.8 22.99.0 14.5 10.7 7.6 10.312.2 7.15-Araf 8.7 7.8

2-Araf trace 1.0 trace 1.0 2.4 trace tracetrace3-Araf trace0.2 trace 3.0 trace tracetrace1.22,5-Araf

3,5-Araf 1.4 1.3 trace 2.5

trace trace 1.2 1.4 traceterminal-Xylp trace 3.1trace 4.2 2.9 1.64-Xylp trace2.7 7.0 3.8 1.2 2.0

3.9 1.9 2.7 0.7 9.92.9 2.7terminal-Gal 3.8 2.2 1.832.8 26.3 22.3 27.7 30.0 29.2 24.0 40.23-Gal 27.628.9trace 1.1 trace 2.1 1.2trace4-Gal1.8 trace 1.6 trace 1.36-Gal tracetrace 2.2 1.3 1.4

22.5 24.9 23.3 29.8 20.620.9 18.93,6-Gal 17.8 26.6 25.7

terminal-Man trace trace trace trace trace4.5 2.4 13.1 1.94-Man 6.2 4.3 4.3 6.8 4.4

4,6-Man 1.4 trace 1.1 0.9

traceterminal-Glc 1.4 3.4trace trace trace trace trace1.1 trace4-Glc trace

An attempt was made to solubilise moregalactomannan by using an 8 M KOH extrac-tion. However, as shown in Table 5, the 8 MKOH extract contained little mannan and waspredominantly arabinogalactan.

The presence of (1�4)-linked glucose moi-eties may be indicative of the presence ofgalactoglucomannans, xyloglucans or gluco-mannans since no starch was detected in thegreen beans.

Linkage analysis of insoluble polymers.—Structural features of the insoluble polysac-charides were investigated by repeatedmethylation of the residue. The main speciesdetected were 4-linked mannose, 3- and 3,6-linked galactose and 4-linked glucose. Typicalresults show a ratio of unsubstituted mannoseto substituted mannose of 20:1, and of 1.7:1for unsubstituted galactose to substitutedgalactose for both species. The average degreeof branching of the galactomannans remain-ing in the residue was similar to that of thematerial extracted with 4 M KOH.

4. Discussion

The results of this study showed no differ-ence in the total polysaccharide content or anymarked difference in the structural features ofthe galactomannans between Robusta ROMand Arabica Caturra. However, differences insolubility, particularly in relation to the ara-binogalactans, did support the idea that thepolysaccharides of Robusta beans were moreeasily extracted than those of Arabica.

Robusta ROM contained amounts of ahighly soluble arabinogalactan which pos-sessed more branch points and more extendedside-chains than arabinogalactans found inArabica Caturra. This may be one of thereasons that the arabinogalactans of Robustawere more easily solubilised than those ofArabica. It is feasible that the accommodationwithin the cell wall matrix of a polysaccharidewith more extended side-chains may producea cell wall with different physicochemicalcharacteristics (a more open matrix?) than onewhich does not possess such polymers. This

M. Fischer et al. / Carbohydrate Research 330 (2001) 93–101100

different architecture could effect not only therelative ease of solubility of the cell wall as awhole, but also its susceptibility to dissolutionby enzymes or solvents.

Both varieties possessed galactomannanswith ratios of unbranched to branched man-nose between 10:1 and 30:1. While the overallstructures did not differ significantly betweenthe two varieties, higher amounts of the water-soluble galactomannan were recovered in Ro-busta ROM and it was more highly branched.The galactomannans which remained in theresidue or were extracted with concentratedalkali had degrees of branching close to 20:1.Except in the water soluble fraction, no evi-dence was found to support the idea thatRobusta contained more branched galac-tomannans than Arabica.10,11 The average de-gree of branching observed was higher thanthe 100:1 value reported by Bradbury andHalliday7,8 although they use harsher extrac-tion procedures.

The main difficulty in characterising themannans or galactomannans which are

present in the coffee bean lies in the highproportion of insoluble polymers, only about1/3 of the total CWM was solubilised by thesequential fractionation process, despite thefact that an alkali concentration of 8 M wasused.

A certain amount of arabinogalactan re-mained in the residue despite its intrinsic solu-bility, pointing to the importance of cell-wallarchitecture in determining the physicochemi-cal properties of its components. This wassupported by the fact that arabinogalactanswith similar structural features were recoveredin all the cell wall fractions, indicating that theimmediate environment of the polysaccharidehad more influence on its solubility than itsbasic structural features.

The complex heterogeneity of the arabino-galactans shown in this study justifies a morein-depth investigation of their structural fea-tures. In particular their association withprotein, acidic monosaccharide content andinteraction with the galactomannans are areaswhich need further study. Type II arabino-

Table 9Glycosyl-linkage analysis of the galactomannan-rich polysaccharide fractions from Arabica and Robusta coffee

Water-soluble 1 M KOH-soluble NaClO2-solubleFraction 4 M KOH-soluble 8 M KOH-soluble

ArabicaRobusta RobustaLinkage ArabicaRobustaArabicaRobustaArabicaRobustaArabica

7.9 9.4 14.0 18.7terminal-Araf 4.5 18.6 17.6 16.9 16.9 16.14.6 9.5 7.6 7.8 3.4 5.25-Araf 7.1 8.11.2

2-Araf -trace 4.9 4.4 9.6 1.7 2.4 trace 3.22.03-Araf 2.3 - 2.6 6.2 trace 2.32.7

-2,5-Araf 0.53,5-Araf 1.4 2.0 trace 1.5 -

7.31.2terminal-Xylp 1.9 5.6 2.2. 3.06.36.98.39.0 2.015.96.62.14-Xylp 1.1

5.3 2.1 15.1 3.8terminal-Gal 3.1 3.5 3.3 2.8 2.9 3.78.5 12.7 15.3 10.4 15.7 17.4 33.2 17.33-Gal 6.2 6.49.3 3.1 2.34-Gal 1.5

trace 1.26-Gal trace 1.2 1.5 2.87.1 7.9 9.8 13.8 11.43,6-Gal 7.8 12.9 14.4 23.4 12.7

3.02.7 4.2terminal-Man 1.12.12.23.31.63.82.148.7 25.0 22.8 19.64-Man 9.641.0 20.0 15.4 14.5 16.2

2.2 trace4,6-Man trace0.81.11.11.81.83.41.7

2.33.12.91.5terminal-Glc5.9 4.34-Glc 1.22.3trace1.95.85.02.9

M. Fischer et al. / Carbohydrate Research 330 (2001) 93–101 101

galactans such as those identified in coffeebeans are often linked to hydroxyproline orserine residues, but so far no such links havebeen identified in coffee.17

To date there has been a general assump-tion that the coffee bean consists of an amal-gam of arabinogalactans, galactomannans andcellulose. The presence of additional and per-haps novel polysaccharides in coffee beans isseldom discussed. The difficulty of solubilisingthe bulk of the cell wall polysaccharides indi-cates an intimate association between some ofthe arabinogalactans, galactomannans andcellulose. While this could be caused by non-covalent interactions between the differentpolysaccharides, it may also indicate the pres-ence of covalent linkages and therefore thepresence of a new class of polysaccharidespossessing, as yet, undetermined structuralfeatures.

Acknowledgements

The authors are grateful to John Rogers forproviding the samples and for helpful discus-sions. We also thank Laurent Fay and Sylvi-ane Metairon for their help with the GC/MSanalysis.

References

1. Thaler, H.; Arneth, W. Z. Lebensm. Unters. Forsch. 1968,138, 26–35.

2. Thaler, H. Z. Lebensm. Unters. Forsch. 1970, 143, 343–349.3. Wolfrom, M. L.; Plunkett, R. A.; Laver, M. L. Agric. Food

Chem. 1960, 8, 58–65.4. Wolfrom, M. L.; Patin, D. L. Agric. Food Chem. 1961, 12,

376–377.5. Wolfrom, M. L.; Plunkett, R. A.; Laver, M. L. J. Org.

Chem. 1961, 26, 4533–4535.6. Wolfrom, M. L.; Patin, D. L. J. Org. Chem. 1963, 30,

4060–4063.7. Bradbury, A. G. W.; Halliday, D. J. 12e Colloque Scien-

tifique International sur le Cafe ; ASIC: Paris, 1987; pp.265–269.

8. Bradbury, A. G. W.; Halliday, D. J. J. Agric. Food Chem.1990, 36, 389–392.

9. Navarini, L.; Gilli, R.; Gombac, V.; Abatangelo, A.; Bosco,M.; Toffanin, R. Carbohydr. Polym. 1999, 40, 71–81.

10. Clifford, M. N. In Coffee Botany, Biochemistry and Produc-tion of Beans and Be6erage ; Clifford, M. N.; Wilson, K. C.,Eds. Chemical and physical aspects of green coffee andcoffee products. Croom Helm: London, 1985; pp. 305–374.

11. Clifford, M. N. Tea Coffee Trade J. 1986, 158, 30–32.12. Trugo, L. C. In Coffee, Chemistry ; Clarke, R. J.; Macrae,

R., Eds. Carbohydrates. Elsevier: Amsterdam, 1985; Vol.1, pp. 83–114.

13. Ciucanu, I.; Kerek, F. Carbohydr. Res. 1984, 131, 209–217.14. Redgwell, R. J.; Fischer, M.; Kendal, E.; MacRae, E. A.

Planta 1997, 203, 174–181.15. Redgwell, R. J.; Melton, L. D.; Brasch, D. J. Plant Physiol.

1992, 98, 71–81.16. Selvendran, R. R.; O’Neill, M. A. Methods Biochem. Anal.

1987, 32, 25–153.17. Clarke, A. E.; Anderson, R. L.; Stone, B. A. Phytochem-

istry 1979, 18, 521–540.

.